I don’t have an explicit theory of how this works; for example, I would consider “pleasing others” in an experience machine meaningless, but “eating a cake” in an experience machine seems just as okay as in real life (maybe even preferable, considering that cakes are unhealthy). A fake memory of “having eaten a cake” would be a bad thing; “making people happier by talking to them” in an experience machine would be intrinsically meaningless, but it might help me improve my actual social skills, which would be valuable. Sometimes I care about the referent being real (the people I would please), sometimes I don’t (the cake I would eat). But it’s not the people/cake distinction per se; for example in case of using fake simulated people to practice social skills, the emphasis is on the skills being real; I would be disappointed if the experience machine merely gave me a fake “feeling of having improved my skills”.
I imagine that for a psychopath everything and everyone is instrumental, so there would be no downside to the experience machine (except for the risk of someone turning it off). But this is just a guess.
I suspect that analyzing “the true preferences” is tricky, because ultimately we are built of atoms, and atoms have no preferences. So the question is whether by focusing on some aspect of the human mind we got better insight to its true nature, or whether we have just eliminated the context that was necessary for it to make sense.
I don’t have an explicit theory of how this works; for example, I would consider “pleasing others” in an experience machine meaningless, but “eating a cake” in an experience machine seems just as okay as in real life (maybe even preferable, considering that cakes are unhealthy). A fake memory of “having eaten a cake” would be a bad thing; “making people happier by talking to them” in an experience machine would be intrinsically meaningless, but it might help me improve my actual social skills, which would be valuable. Sometimes I care about the referent being real (the people I would please), sometimes I don’t (the cake I would eat). But it’s not the people/cake distinction per se; for example in case of using fake simulated people to practice social skills, the emphasis is on the skills being real; I would be disappointed if the experience machine merely gave me a fake “feeling of having improved my skills”.
I imagine that for a psychopath everything and everyone is instrumental, so there would be no downside to the experience machine (except for the risk of someone turning it off). But this is just a guess.
I suspect that analyzing “the true preferences” is tricky, because ultimately we are built of atoms, and atoms have no preferences. So the question is whether by focusing on some aspect of the human mind we got better insight to its true nature, or whether we have just eliminated the context that was necessary for it to make sense.